lockjaw encodes a zebrafish tfap2a required for early neural crest development

The vertebrate neural crest forms the pigmentation of the body, peripheral innervation, and much of the cartilage and bone of the skull. Disruption of neural crest development in humans results in neuropathies, aberrant pigmentation and craniofacial defects(Schilling, 1997; Maschhoff and Baldwin, 2000; Epstein, 2001; Manie et al., 2001). Molecular mechanisms that control the differentiation of many of these derivatives have been studied extensively, but less is known about the early events that induce neural crest cells and specify their fates. Neural crest cells are induced at the lateral edges of the neural plate during neurulation, lose adhesion with the neuroepithelium and migrate along defined pathways(LeDouarin, 1982; Christiansen et al., 2000). These early crest cells are thought to be multipotent, and only later acquire their fates from their migratory environments, but a growing body of evidence suggests that many crest cells may be specified much earlier.

Two secreted signals have been suggested to induce neural crest during gastrulation, bone morphogenetic proteins (Bmps) and Wnts(Nieto, 2001; Aybar and Mayor, 2002). One key step is thought to be attenuation of Bmp signaling in the neural plate by specific antagonists (Chordin, Noggin) and posteriorizing signals (Wnt, Fgf and Retinoic acid) from adjacent mesoderm. Recent evidence has also supported the role of Wnt expressed in the adjacent ectoderm as a neural crest inducer(Garcia-Castro et al., 2002),and Delta/Notch signaling in the specification of trunk neural crest(Cornell and Eisen, 2000; Cornell and Eisen, 2002; Endo et al., 2002). Following its induction, the location of a neural crest cell in the ectoderm prior to migration then plays a crucial role in determining its segment and cell-type specific fates. Cranial neural crest cells that will migrate into the pharyngeal arches are organized along the anteroposterior axis into regions fated to form each arch segment (Lumsden et al., 1991; Schilling and Kimmel, 1994; Kontges and Lumsden, 1996). In zebrafish, the position within the mediolateral(ML) axis of the premigratory neural crest also determines if a cell will adopt a pigment cell fate over a neural or glial fate(Schilling and Kimmel, 1994; Kelsh and Raible, 2002). Wnt signals arising from the dorsal midline of the ectoderm bias this choice of cell type; activation of Wnt signaling promotes pigment cell fates(Dorsky et al., 1998). Thus,the same signals that induce neural crest cells may also specify their future migration patterns and fates within the embryo.

In addition to signaling molecules, a growing list of transcription factors appear to function in the early premigratory crest. These include the related zinc-finger genes snail, slug(Nieto et al., 1994; Sefton et al., 1998; del Barrio and Nieto, 2002)and twist (Soo et al.,2002), the forkhead transcription factor foxd3(Dottori et al., 2001; Kos et al., 2001; Sasai et al., 2001), Zic genes(Nakata et al., 1997; Nakata et al., 1998), and the paired transcription factor pax3(Epstein et al., 1991; Tassabehji et al., 1993). Another family of transcription factors implicated in early neural crest development is the Activator Protein 2 (AP2) family, defined by a unique,highly conserved DNA-binding domain(Hilger-Eversheim et al.,2000). There are at least four AP2 genes in mice (Tcfap2a,Tcfap2b, Tcfap2g and Tcfap2d), three of which are co-expressed in the early ectoderm and neural crest(Mitchell et al., 1991; Chazaud et al., 1996; Moser et al., 1995; Moser et al., 1997; Werling and Schorle, 2002). Mice mutant for AP2 family genes reveal important functions in development,for example Tcfap2a mutant mice have craniofacial, neural tube, body wall, limb and eye defects (Schorle et al., 1996; Zhang et al.,1996; Morriss-Kay,1996). Chimaeric analyses using Tcfap2a mutant mice show that these five classes of defects are independent of one another, implicating Tcfap2a in multiple processes of development(Nottoli et al., 1998). The craniofacial defects in Tcfap2a^–/– mice are correlated with extensive cell death in the cranial neural crest and neural tube. Neural crest derived pharyngeal cartilages and middle ear bones, as well as cardiac neural crest of the heart outflow tract are malformed and reduced(Brewer et al., 2002). However,two early cranial crest markers, Pax3 and Twist, show no disruption in these mutant mice and neural crest migration appears to occur normally, as shown by combining a Wnt1-lacZ transgene with the Tcfap2a^–/– mutation. This implicates Tcfap2a in the development of several neural crest populations, yet relatively little is known about the roles of Tcfap2 at early stages of neural crest development.

We show that the craniofacial mutant lockjaw (low)disrupts a zebrafish tfap2a that is required for the formation of subsets of neural crest-derived cell types (subsets of sensory neurons,cartilages and pigment cells) and pharyngeal segments (second pharyngeal arch). Zebrafish tfap2a is expressed in the non-neural ectoderm during gastrulation and then is upregulated in premigratory neural crest. We find that low is required at these early stages for the survival of neural crest migrating from the hindbrain and spinal cord. This is reflected in disruption and loss of early gene expression in the neural crest, including that of tfap2a itself. We focus on the subpopulations of neural crest disrupted in low mutants and show that the requirement for low for neural crest cell survival is cell autonomous and segment specific. This is the first evidence for such segment and cell-type-restricted functions of tfap2a in neural crest.

Embryos were obtained in natural crosses and staged according to Kimmel et al. (Kimmel et al., 1995). The low^ts213 mutation was identified in an ENU screen for abnormal jaw morphology (Schilling et al.,1996b), maintained in a Tubingen (Tu) background and heterozygotes were outcrossed to a reference (WIK) strain for mapping. Cartilages were stained with Alcian Blue, dissected and flat-mounted(Kimmel et al., 1998). To label aortic arches, fluorescent microspheres (Molecular Probes) were microinjected into the axial vein at 72 hours postfertilization (hpf) (0.1μm diameter) (Isogai et al.,2001).

The low^ts213mutation was mapped using bulked segregant analysis of F2 embryos and genome scanning with SSLPs(Geisler, 2002). A candidate gene for low^ts213, tfap2a, mapped within 6 cM of our map position for low^ts213 as does the montblanc (mob^m819) mutation (W. Driever,personal communication). tfap2a was cloned from low mutants and wild-type siblings using primers designed against the published sequence(GenBank: AF457191). The fifth exon of tfap2a was amplified from genomic DNA preparations of individual low mutants and wild-type siblings using primers designed against flanking intron sequence (forward:CGCTCAGGTCTTATAAATAGGC and reverse: CTGAGAGGTGGCTATTTCCCG). Restriction enzyme digest of PCR-amplified exon five by BlpI (New England Biolabs) was used to identify low mutants and heterozygotes from a panel of 192 mutant and wild-type siblings. Sequencing of DNA was performed using the BigDye Terminator Sequencing reagent on a thermal sequencer and reactions run on an ABI PRISM 310 sequencer (PE Applied Biosystems).

PAC DNA clones (library obtained from C. Amemiya) containing tfap2a were identified by PCR directed to exon five (as above) and confirmed using Southern analysis (Amemiya et al., 1999). DNA purified for injection was diluted to a concentration of 1ug/μl and ∼10 nl were injected into lowmutants at the one- to two-cell stage(Fishman et al., 1997). Rescue was defined by the presence of small numbers of melanophores at 24 hpf, which are never seen in uninjected low mutant embryos.

Wild-type and low^ts213 mutant embryos were injected with 10 kDa DMNB-caged fluorescein (5 mg/ml). Fluorescein was then activated when embryos reached the eight-somite stage along the dorsal and lateral neural keel (David et al.,2002). Labeled cells were then inspected at 28 hpf either visually or by immunodetection of fluorescein. Apoptotic cell death was detected in whole embryos by terminal transferase dUTP nick-end labeling (TUNEL) using modifications suggested by the manufacturer (In Situ Cell Death Detection Kit;POD; Roche). Embryos were fixed for 3-4 hours at room temperature in 4%paraformaldehyde, permeabilized in acetone and blocked in 2% goat serum for 2-3 hours at room temperature.

Wild-type and mutant donor embryos were injected at the one- to two-cell stage with a mixture of 3% TRITC-dextran (neutral, 10,000 M_r) and 3% biotinylated-dextran (lysine-fixable, 10,000 M_r). At the three-somite stage, labeled premigratory neural crest cells were transplanted into low mutant hosts as described previously (Schilling et al.,1996a), into segment-specific locations according to the fate map(Schilling and Kimmel, 1994). Hindbrain cells are often co-transplanted with neural crest cells in these experiments, conveniently marking their segments of origin.

Three different morpholino antisense oligonucleotides (translation start site of tfap2a1, ATGmo – ATTTCCAAAGCATTTTCATTGGTTG; intron 3 splice donor junction; 3.1mo – GAAATTGCTTACCTTTTTTGATTAC; intron 5 splice acceptor junction, 5.1mo – CCTCCATTCTTAGATTTGGCCCTAT) were obtained from Gene Tools (Philomath, OR) and injected at the one- to two-cell stage. Efficacy of morpholinos directed against splice sites was evaluated using RT-PCR (Fig. 2F) with forward primers tfap2a-3f and tfap2a-4f designed to 464 bp and 777 bp of tfap2a1 (GenBank Accession Number AF457191) (GGTCACGGCATTGATACTGG;CTGAGTGCCTGAACGCTTCC) in conjunction with reverse primer tfap2a-3r designed to 1184 bp of tfap2a1 (GTCAAACAGCTCTGAATCCC).

In situ hybridization was carried out as described previously(Thisse et al., 1993). An antisense 2050 bp riboprobe for tfap2a(Furthauer et al., 1997) was generated using SP6 polymerase from RZPD clone: MPMGp609H2035Q8(www.rzpd.de)following linearization by KpnI. Probes and antibodies used were: crestin (Rubinstein et al.,2000; Luo et al.,2001), dlx2 (Akimenko et al., 1994), endrb(Parichy et al., 2000b), foxd3 (Odenthal and Nusslein-Volhard, 1998), dct(Kelsh et al., 2000), fms and xdh (Parichy et al., 2000a), gch(Pelletier et al., 2001), mitfa (Lister et al.,1999), hoxa2 and hoxb2(Prince et al., 1998; Amores et al., 1998), gsc (Stachel et al.,1993), sox9b (Li et al., 2002), isl1(Inoue et al., 1994), spa/kit (Parichy et al.,1999), sna2 (Thisse et al., 1995), anti-Hu, mjAb 16A11(Marusich et al., 1994), and anti-TH (Guo et al.,1999).

Zebrafish embryos homozygous for the recessive lethal lockjaw(low^ts213) allele have reduced body pigmentation and cartilages of the pharyngeal skeleton. In wild-type embryos, 40-50 melanocytes differentiate on each side of the head at 28 hpf, just anterior and posterior to the otic vesicles (Fig. 1A),but embryos homozygous for low^ts213 lack most of these early melanocytes, while formation of the pigmented epithelium of the retina is unaffected (Fig. 1B). Older embryos and larvae can be identified by ∼30% reduction in pigment cell number, small eyes and a ventrally protruding jaw(Fig. 1C). Cartilage differentiates in the jaw and pharyngeal arches of wild-type larvae at 72 hpf(Fig. 1D,F), but low^ts213 mutant cartilages are severely reduced,particularly in the second or hyoid arch(Schilling et al., 1996b)(Fig. 1E,G). Alcian staining revealed that the dorsal hyosymplectic cartilage of the hyoid is consistently absent in low^ts213 mutants, while the ventral ceratohyal is reduced (Fig. 1G). By contrast, the jaw or mandibular arch is much less affected in low^ts213 mutants. Its dorsal element, the palatoquadrate,forms an ectopic contact with and fuses to the anterior basicranial commissure of the skull (Fig. 1G,I). By dissecting these cartilages free of other tissues we find that branchial cartilages are present in over 95% of low mutants though variably reduced (Fig. 1E), while severe hyoid defects occur in all cases. In addition, other pharyngeal cell types such as the aortic arches are also reduced in low mutants as revealed by injection of fluorescent microspheres into the bloodstream(Fig. 1J-L). These defects in pigment and pharyngeal arches suggest that low may be required in their common precursors in the neural crest.

Bulked segregant analysis of an F2 mapping cross localized low to linkage group 24 (LG24) on a meiotic map between markers z23011 and z3399(Fig. 2A). The transcription factor tfap2a maps between these markers and is expressed in neural crest (B. Thisse, S. Pflumio, M. Furthauer, B. Loppin, V. Heyer, A. Degrave,R. Woehl, A. Lux, T. Steffan, X. Q. Charbonnier and C. Thisse, Zebrafish Information Network, http://zfin.org,2001), therefore we tested tfap2a as a candidate gene for low. The amino acid sequence shows high similarity to mouse Tcfap2a (68.5% overall identity, 98.3% identity within the DNA-binding and dimerization domains; data not shown). Further comparison of the corresponding genomic sequences revealed conservation of the exon-intron structure between the two species, indicating that this is a zebrafish orthologue of mouse Tcfap2a. Consequent sequence analysis of tfap2a in low^ts213 mutant larvae revealed a point mutation of an adenosine (at 900 bp of the tfap2a1 sequence) to a thymidine, creating a stop codon in place of a highly conserved lysine in the DNA-binding domain (Fig. 2B). This truncates the protein coding sequence and would result in deletion of the C terminus of the protein, including both the dimerization domain and most of the DNA-binding domain (Williams and Tijan, 1991; Garcia et al., 2000). This unique change also creates a restriction site for BlpI, allowing identification of low homozygotes by PCR amplification of exon5(Fig. 2C). Low fails to complement montblanc^m819 which is also a mutation in tfap2a and disrupts development of catecholaminergic neurons(Neuhauss et al., 1996) (W. Driever, personal communication)

As a direct demonstration that tfap2a is responsible for the low phenotype we performed rescue experiments. We took advantage of the consistent absence of melanophores in low^ts213 mutant embryos between 24-27 hpf. Microinjection of 5-10 ng of a PAC (19:226)containing the tfap2a gene into embryos at the one- to two-cell stage, rescued local patches of melanophores in 28% (14/50) of lowmutants, or in some cases gave unilateral rescue over large areas(Fig. 2D), while control PAC(D17:5) injections failed to do so (Fig. 2D).

To show further that reduction in tfap2a function causes the low phenotype, we injected one- to two-cell stage wild-type embryos with morpholino antisense oligonucleotides targeted to tfap2a(Fig. 2E-G). We first injected a morpholino directed against the putative translation start site of the tfap2a1 splice form (ATGmo) but as much as 10 ng/embryo had no specific effect on neural crest derived pigment or cartilage. In zebrafish,there are multiple splice forms of tfap2a, with three alternative first exons as in mouse (Meier et al.,1995). Therefore, we tried morpholinos targeted to a splice donor site at the third exon (3.1mo) and to a splice acceptor site at the fifth exon(5.1mo), at positions 537 and 819 of the tfap2a1 mRNA sequence,respectively (Fig. 2E). Five nanograms/embryo of the 5.1mo alone was sufficient to effectively phenocopy the pigment and jaw defects characteristic of low mutants(Fig. 2G; Table 1). This phenotype correlated with a dramatic change in tfap2a transcripts, in which regions of adjoining introns are not spliced correctly, as detected by RT-PCR using primers that amplify over these splice sites(Fig. 2F). Microinjection of the 3.1mo caused only a subtle delay in pigmentation when injected at 5 ng/embryo, but more effectively reduced pigmentation at 10-20 ng/embryo concentrations and this correlated with weaker affects on tfap2a RNA splicing (Fig. 2F). Notably,morpholino injections did not increase the severity of the phenotype when injected into homozygous low^ts213 mutants, supporting the notion that low^ts213 is a complete loss-of-function mutation.

Zebrafish tfap2a is first expressed at late blastula and early gastrula stages in a broad ventral domain of ectoderm(Fig. 3A,B), which corresponds to the non-neural ectoderm as revealed by comparison with gata3(Neave et al., 1995). By 80%epiboly, expression is upregulated in putative progenitors of the neural crest, and subsequently becomes restricted to two thin bands within the ectoderm, stretching from approximately the level of the prospective forebrain-midbrain junction to the posterior hindbrain initially and then progressively expanding posteriorly to spinal levels(Fig. 3C,D). Outside the ectoderm, tfap2a is also expressed in intermediate/lateral plate mesoderm (Fig. 3L). tfap2a expression in neural crest only partially overlaps the expression of foxd3 and sna2 at the six-somite stage,revealing early differences in gene expression within different populations of neural crest (data not shown).

By 24 hpf, tfap2a expression becomes confined predominantly to migrating neural crest and neurons within the central nervous system (CNS)(Fig. 3E). No expression can be detected in the epidermis at this stage. Expression in the pharyngeal arches is widespread throughout the neural crest-derived mesenchyme(Fig. 3F) and in small groups of cells migrating medial to each somite (data not shown). tfap2a is also expressed in neuronal groups within the telencephalon and ventral mesencephalon (Fig. 3G), as well as segmental groups of neurons within each rhombomere of the hindbrain(Fig. 3H). Prospective Rohon-Beard (RB) neurons within the spinal cord, which originate alongside neural crest, also express tfap2a at 24 hpf(Fig. 3I)(Cornell and Eisen, 2000).

To determine if expression of tfap2a mRNA was disrupted in low mutants we performed in situ hybridization at 12 and 24 hpf. Although in wild-type controls, dark labeling of tfap2a-expressing cells was achieved with coloration reactions lasting only a few hours, in low mutants incubation overnight gave very weak labeling(Fig. 3J,K). These results suggest either that the mutant mRNA is unstable and degraded after transcription or, alternatively, that tfap2a function is required for its own expression.

To determine the stage at which neural crest development is disrupted in low mutants, we analyzed the expression of molecular markers of premigratory and migrating crest. Cranial neural crest is first visible morphologically at 10 hpf and becomes progressively more prominent posteriorly during somitogenesis. In six-somite stage wild-type embryos, the forkhead transcription factor foxd3 is expressed throughout neural crest cells emerging from midbrain, hindbrain and anterior spinal regions, while crestin (ctn) is strongly expressed only in neural crest posterior to the otic vesicle (Fig. 4A,C,E). In low mutants, ctn expression is virtually eliminated, while, surprisingly, foxd3 expression is reduced in hindbrain neural crest, but not in more anterior neural crest adjacent to the midbrain or at more posterior somite levels(Fig. 4B,D,F). Furthermore,several other genes expressed prior to this stage in the premigratory neural crest including snail2 (sna2) and sox9b are expressed in the correct locations in low mutants (data not shown). By mid-migratory stages (14-18 hpf), there are dramatic reductions in sox9b (Fig. 4G,H), but sna2 remains relatively unaffected, particularly when compared by two-color in situ hybridization with defects in ctn expression(Fig. 4I-L). These results suggest that there are distinct requirements for tfap2a in specification of subpopulations within the crest.

Consistent with an early role in neural crest cell specification, markers of pigment precursors are disrupted in low mutants. In 26 hpf wild-type embryos, both dopachrome tautomerase (dct) and kit are expressed in melanocyte precursors just anterior and posterior to the otic vesicle, dct is also strongly expressed in the pigmented epithelium of the eye (Fig. 5A,C). In low mutants, expression of dct is reduced to a small cluster of post-otic crest cells at this stage, but is expressed normally in the eye (Fig. 5B), and kit expression is severely reduced(Fig. 5D). This is consistent with the established function of Tfap2a in regulating kit in other vertebrates (Huang et al.,1998). Likewise, gch which is expressed in both melanophore and xanthophore precursors(Fig. 5E) marks many fewer cells in low mutants at 24 hpf(Fig. 5F) particularly in the vicinity of the otic vesicle. We found reductions in expression of several other pigment precursor markers at this stage including mitfa, endrb,xdh and fms, and reductions in differentiated xanthophores and iridophores at 72 hpf (Lister,2002) (data not shown), demonstrating that a proportion of all pigment cell lineages requires tfap2a function.

Neural crest cells in the trunk form in part through lateral inhibition of neurogenesis by neuroblasts that form dorsal sensory neurons (the RB cells) in the spinal cord (Cornell and Eisen,2000; Cornell and Eisen,2002). RB cells as well as crest cells express tfap2a.(Fig. 3I). RB cells form in low mutants, though their numbers are slightly reduced(Fig. 5I,J). In 12 hpf wild-type embryos, the transcription factor Islet1 (Isl1) marks 12-14 neural precursors in each of the trigeminal placodes(Fig. 5K), which are derived from epidermal placodes. The number is reduced to five or six cells in low mutants (Fig. 5L). By 36 hpf the low mutant trigeminal ganglion is well-formed but contains only 60-70% of the normal number of neurons(Fig. 5M,N). Other cranial ganglia, including both neural crest-derived neurons of the facial (VII),glossopharyngeal (IX) and vagal (X) nerves, as well as placodally derived epibranchial ganglia are also reduced in size and neuron number. Somewhat surprisingly, neural crest-derived glial cells in these ganglia, as assayed by expression of foxd3 at 28 hpf are only mildly reduced in number, by∼10% (Fig. 5G,H). By 72 hpf, mutants lack enteric neurons derived from neural crest(Fig. 5Q,R), as well as sympathetic neurons (data not shown), while dorsal root ganglia are less affected (Fig. 5O,P). Tyrosine hydroxylase (TH)-positive neurons of the carotid bodies, which form just ventral to the hyoid arch, are also reduced in low(Fig. 5S,T), as are other TH-positive neurons in the CNS (data not shown). Thus, neuronal defects in low mutants include a variety of sensory neurons derived from both neural crest, and epidermal placodes, but spare many sensory neurons and glia of cranial and spinal ganglia.

The changes in gene expression in the neural crest of low could reflect a downregulation of target gene expression or, alternatively, a failure of neural crest cells to migrate, proliferate or survive. To distinguish between these possibilities, we followed cranial neural crest morphogenesis in living low mutant embryos(Fig. 6A-D). Morphological inspection with Nomarski optics at 24 hpf revealed the presence of reduced arch primordia within which cells appear disorganized and not arranged in linear rows as in wild type (data not shown). To examine the segmental origins of these disorganized arches, we followed labeled neural crest cells in low mutants using the fate map for premigratory neural crest generated for 12 hpf embryos (Schilling and Kimmel, 1994). The photoactivatable dye DMNB-caged fluorescein was injected into the progeny of low heterozygotes at the one-cell stage and the dye was UV-activated in a local region of the cranial neuroepithelium when embryos reached the eight- to ten-somite stage. This technique typically labeled both neural tube and neural crest cells, marking the segmental origins of the crest cells. In both wild-type and lowmutant embryos, fluorescent cells were found to migrate to the appropriate pharyngeal arch. For example, photoactivation in the region immediately posterior to the eye labels stream I and later the mandibular arch(Fig. 6A,B), while labeling just anterior or posterior to the otic vesicle labels stream II and the hyoid or stream III and the branchial arches, respectively(Fig. 6A-D). These results suggest that low mutant neural crest cells migrate and roughly populate the appropriate segments along the anteroposterior axis.

To determine if low is required for neural crest cell survival, we analyzed cell death patterns in wild-type and low homozygous embryos using TUNEL analysis (Fig. 6E-G). Strikingly, although fewer than 10 TUNEL-positive cells were typically observed in wild-type embryos, by the six-somite stage, 30-90 TUNEL-positive cells/embryo were observed in the dorsal ectoderm at all anteroposterior levels in 25% of embryos derived from an intercross between low heterozygotes (Fig. 6E,F). At slightly earlier stages, labeling was restricted to midbrain/hindbrain levels, correlated with the timing of neural crest emigration (data not shown). Furthermore, by the 12-somite stage, the number of dying cells in the cranial region had decreased, and labeling spread to spinal levels (Fig. 6I). TUNEL labeling subsequently became less pronounced by late segmentation stages. Some TUNEL-positive cells at the six-somite stage were identified as neural crest cells by colocalization with the few remaining ctn-positive cells in low, and this further confirmed that the death occurs specifically in low mutants (Fig. 6G,H). Thus at least some neural crest cells require low(tfap2a) function for their survival, and this requirement appears to be limited to a short early period of neural crest development.

The mandibular arch is much less affected in low mutants than other pharyngeal arches, and in some cases the first and second arch cartilages appear to fuse, reminiscent of a homeotic transformation of the second arch into a first arch identity(Fig. 1)(Hunter and Prince, 2002). In wild-type embryos at 32 hpf, dlx2 is strongly expressed in five out of the eventual seven pharyngeal arches as well as in the forebrain(Fig. 7A,C). In lowmutants, only expression in the mandibular and weak expression in some branchial arches can be detected, yet the mandibular expression domain is approximately the normal size and shape(Fig. 7B,D). Mandibular neural crest is Hox negative, in contrast to crest migrating into the hyoid arch,which expresses Hox group 2 genes. To determine if the pharyngeal arch defects in low are due to disruption of Hox gene expression we examined the expression of hoxa2 and hoxb2 at 28 hpf. Both are dramatically reduced in the hyoid arch in low mutants, but not in the hindbrain (Fig. 7E,F). This is similar to results from analyses of the enhancer elements driving mouse Hoxa2 expression in the hyoid neural crest, which are regulated by AP2 genes (Maconochie et al.,1999). Defects in the hyoid arch are most severe dorsally with loss of the hyosymplectic and fusion of its remnant with the mandibular skeleton. This correlates with a loss of gsc expression in the dorsal hyoid at 36 hpf (Fig. 7G-J)(Miller at al., 2000). A loss of Hox2 group genes in the second arch, coupled with the loss of dlx2and gsc could account for some of the segmental fusions and abnormal jaw articulations we observe in low mutants.

Tcfap2a is expressed throughout the ectoderm and required for neural tube closure in mice (Schorle et al., 1996; Zhang et al.,1996) as well as playing important roles in epidermis development in both Xenopus and mice(Pfisterer et al., 2002; Luo et al., 2002). Thus,neural crest defects in low mutants might be caused indirectly by defects in other ectodermal derivatives. To exclude this possibility, we used mosaic analysis to test the capacity of wild-type cells to rescue the mutant phenotype when surgically transplanted into the second pharyngeal arch(Fig. 8). Premigratory neural crest that forms this arch has been mapped just anterior to the otic vesicle at 12 hpf (Schilling and Kimmel,1994). Cranial neural crest cells were transplanted from this location in a biotin-labeled donor at 12 hpf into the same region on one side of the neural crest of an unlabeled host and their fates followed until 72 hpf(n=24; Fig.8 A-C) (see Schilling et al., 1996b). The untransplanted side served as a control and allowed unambiguous identification of the host as mutant or wild type. Wild-type cells transplanted into wild-type hosts often contributed neurons to rhombomeres 4 and 5 of the hindbrain as well as a stream of neural crest within the hyoid arch(Fig. 8A-A″). Similar transplants from wild-type donors into low mutant hosts also resulted in large numbers of crest cells in the hyoid(Fig. 8B-B″), and in some cases this also restored cartilage and muscle differentiation in the hyoid arch (Fig. 8B′′′), revealing non-autonomous requirements for low in mesodermally derived muscles that require interactions with skeletogenic crest cells (Noden,1983; Schilling et al.,1996a). By contrast, low mutant cells transplanted into wild-type hosts often failed to migrate and never formed cartilage(n=12; Fig. 8C,C′). These results demonstrate that low acts cell autonomously in the cranial neural crest.

In this study, we show that lockjaw (low) is due to a mutation in a zebrafish AP2 gene, orthologous to mammalian Tcfap2a,and is required for early patterning of the neural crest. The lesion in low is a point mutation in tfap2a that introduces a stop codon in the highly conserved DNA-binding domain, and is predicted to cause a complete loss of function, based on studies in cell culture(Garcia et al., 2000). Consistent with this, injection of two antisense morpholino oligonucleotides directed against splicing sites in tfap2a phenocopies lowbut does not give a more severe phenotype, indicating that low is a complete loss of tfap2a function. Disruption of Tcfap2afunction in mice causes defects in the neural crest-derived craniofacial skeleton and cranial ganglia (Schorle et al., 1996; Zhang et al.,1996). Our analysis of low reveals a conserved role for tfap2a in vertebrates in patterning the cranial neural crest and has implications for understanding human craniofacial and deafness syndromes. We propose a model in which tfap2a is required for (1) early patterning of subpopulations of the neural crest and (2) segment-specific patterning of the hyoid arch (Fig. 9).

In the Tcfap2a mutant mouse, extensive cell death in the neural folds and in the mesenchyme of the pharyngeal arches has been suggested to cause the craniofacial phenotype (Schorle et al., 1996). We also observe increased apoptosis in the cranial region of low at early segmentation stages, indicating a conserved function of tfap2a in cell survival and suggesting a cause for the low phenotype.

This function in preventing apoptosis is conserved between other vertebrate and Drosophila AP2 genes (Kerber et al., 2001). AP2 proteins form a complex with retinoblastoma protein to activate transcription of Bcl2 in mammalian epithelial cells, which directly prevents apoptosis(Decary et al., 2002) and can also prevent apoptosis through inhibition of DNA binding by the cell-cycle progression factor Myc (Hilger-Eversheim et al., 2000). These established roles for AP2 proteins in promoting cell survival suggest that the low phenotype may be due to increased neural crest cell death following loss of tfap2a function. In support of this, we found that some dying cells in low express ctn, a neural crest-specific marker. Apoptosis occurs throughout the neural crest and in the dorsal epidermis, but occurs only during a brief interval of development, between 12-15 hpf, which might account for the partial loss of neural crest derivatives. These results suggest that the low phenotype might be explained by apoptosis.

An alternative hypothesis for the low phenotype is that some neural crest is not specified correctly early in development, resulting in an absence of a subset of pigment and skeletal cells. In support of this, a number of early neural crest markers are reduced or absent in premigratory cranial neural crest cells of low, but expression of Sna2(orthologous to mammalian Snail), is unaffected in the same cells. The expression of Snail2 is also unaffected at stages in which we observe most apoptosis in putative neural crest, suggesting that many neural crest cells remain in low. Given the very early role of the Snail gene family in neural crest specification and migration in Xenopus and chick (LaBonne and Bronner-Fraser, 1998; del Barrio et al., 2002) (reviewed by Nieto, 2002), this suggests that neural crest cells are specified in low, but later patterning is disrupted, possibly leading to apoptosis. These early expression defects of some genes in the neural crest of low correlate with later differentiation of a subset of neural crest derivatives. Thus, in this model we would argue that the loss of certain subpopulations of neural crest in low is a direct result of early patterning defects. This model does not preclude death of these cells by apoptosis at later stages, although we do not have direct evidence that the gene expression defects and patterns of apoptosis are related. Rather apoptosis, when it occurs, appears widespread in low and not confined to any particular regions of the head, as might be expected if death accounted for the regional defects in neural crest in the arches.

Neural crest cells form many different cell types after migration from the neural tube, and interactions between these migrating cells and their environment are known to regulate crest cell fates(LeDouarin, 1982; Noden, 1983; Graham and Smith, 2001). In order to show whether the pigment, neuronal and skeletal defects in low are due to a disruption in signals from their environment or a result of a cell intrinsic defect in the neural crest, we performed transplants of neural crest between wild type and low. These transplants show that tfap2a is required in a cell-autonomous manner for at least a subset of neural crest to migrate and contribute to the pharyngeal arches. In some cases, wild-type crest cells alone rescued surrounding pharyngeal muscles in low, revealing interactions between skeletogenic crest and surrounding mesodermally derived muscles(Noden, 1983; Schilling et al., 1996a). Transplanted neural crest cells taken from low mutants showed reduced migration into wild-type arches, yet many of these cells appeared to migrate normally. This was supported by photoactivation of fluorescein in premigratory neural crest cells in low, many of which then migrated normally into the pharyngeal arches. These results do not resolve which particular subset of neural crest is affected in low, an issue that will require lineage analyses of individual crest cells, but they do suggest that tfap2afunction is only required in some of these cells.

In zebrafish, the most lateral premigratory neural crest cells that migrate early form neurons, in contrast to more medial cells (which will form pigment cells and cartilage) (Schilling and Kimmel, 1994; Raible and Eisen, 1994). One explanation for the differential affects on neural crest-derived cells in low, is that tfap2a is only required during a short period of crest development,perhaps only in early or late migrating cells. Supporting this hypothesis,neural crest-derived neurons and glia of the dorsal root and cranial ganglia are only mildly disrupted in low, in contrast to pigment cells and craniofacial cartilages which are severely reduced.

Even among pigment or cartilage derived from neural crest, only subpopulations appear to require low function, suggesting a more restricted role of tfap2a. For example, melanocytes are heavily reduced in low, in contrast to xanthophores or iridophores, which are less affected. Consistent with this, genes expressed exclusively in melanoblasts (kit, dct) are more severely affected in lowthan others reported to be expressed in both melanoblasts and xanthophores(gch) (Parichy et al.,1999; Kelsh et al.,2000; Pelletier et al.,2001). In mammals, the proto-oncogene Kit is directly regulated by AP2 genes, and in zebrafish a homolog of this gene has been shown to have a crucial role in melanoblast development(Baldi et al., 2001; Parichy et al., 1999). Our data point to a conserved function of tfap2a. in melanocyte development through the regulation of kit and implicate tfap2a in specifying subsets of neural crest-derived pigment cells.

Not all neural crest is affected to the same extent along the anteroposterior axis in low. For example, the cartilages of the mandibular arch are well formed in mutants while other pharyngeal cartilages are reduced or absent. Likewise, the pattern of foxd3 expression appears to be unaffected in neural crest at midbrain levels that contribute to the mandibular arch in low, but strongly reduced in more posterior neural crest that forms other arches. ctn expression is only mildly reduced in the trunk in mutants but absent from cranial neural crest migrating adjacent to the otic vesicle. Both neural crest of the ectomesenchymal lineage(which forms cartilage) and pigment precursors are affected at this position,as shown by a striking loss of all gch+ pigment precursors at the level of the otic vesicle, as well as loss of sox9b- and dlx2-expressing presumptive skeletogenic cells.

The neural crest that arises immediately adjacent to the otic vesicle, in rhombomeres 4 and 5, migrates anterior to the otic vesicle into the hyoid arch, whereas neural crest from more posterior rhombomeres migrates posterior to the otic vesicle into the branchial arches(Schilling and Kimmel, 1994). The regional defects in neural crest expression during their migration into the arches of low, suggest that tfap2a is required for patterning of a population of neural crest which arise from a restricted region of the hindbrain. This correlates with segment-specific defects in pharyngeal cartilages; while the hyoid arch is severely disrupted in all cases, the mandibular arch is only slightly reduced. More posterior branchial arches are consistently present but reduced in low.

One significant feature of neural crest migrating into the hyoid and more posterior arches, is that they express Hox group 2 genes, in contrast to neural crest that forms the mandibular arch(Hunt et al., 1991; Prince and Lumsden, 1994). Hoxa2, in particular, has been shown to be required for the segmental identity of the hyoid arch, which undergoes a partial transformation to a mandibular morphology in Hoxa2 mutant mice(Rijli et al., 1993; Gendron-Maguire et al., 1993). In low, both hoxa2 and hoxb2 expression are severely reduced in the hyoid and branchial arches, suggesting that tfap2a regulation of Hox genes may be an essential requirement in correct patterning of the cranial neural crest. Although we have not found clear evidence for segmental transformations, as observed in Hoxa2mutant mice, some of the arch fusions and abnormal jaw articulations observed in low may occur as a consequence of loss of expression of Hox2 group genes. Furthermore, we have not detected defects in expression of hoxa2 or hoxb2 in the hindbrain or in patterning of rhombomeres in low, consistent with an independent regulation in the CNS and in neural crest (Prince and Lumsden, 1994; Maconochie et al., 1999; Tumpel et al.,2002). Analysis of enhancer elements in mouse Hoxa2 has shown a requirement for AP2 genes in regulating expression of Hoxa2in hyoid and branchial arch neural crest(Maconochie et al., 1999). Our results indicate that regulation of Hox group 2 gene expression in neural crest is a conserved function of tfap2a in vertebrates, and we propose that the defects in the craniofacial skeleton of low are a direct result of loss of Hox gene function.

However, this loss of Hox group 2 expression is not complete, as shown by reduced ventral expression of hoxa2 in the hyoid and branchial arches, but appears to be specific to the dorsal arch. This dorsal specific defect is even more apparent in the loss of gsc expression specifically in the dorsal hyoid arch. This loss of dorsal hyoid arch expression may account for the cartilage phenotype in low, in which the dorsal element of the hyoid, the hyposymplectic is absent. Interestingly,this segmental restriction of tfap2a function in patterning the hyoid, may also be conserved in vertebrates, as it has been reported that in a mouse Tcfap2a null background, ventrally restricted expression of a Hoxa2 enhancer construct is still present(Maconochie et al., 1999). It is interesting to speculate as to how dorsoventral specificity arose in the pharyngeal arches, as the putative vertebrate ancestor is proposed to have had unjointed arches that show no regional differences along the DV axis, as evidenced for an agnathan, lamprey (de Beer, 1937). Independent regulation of patterning genes in the dorsal and ventral arch may have allowed the innovation of novel vertebrate structures, such as jaws.

Neural crest originated after the divergence of cephalochordates and vertebrates, commensurate with an increase in genome complexity and gene number (Shimeld and Holland,2000). AP2 expression in neural crest is basal to the vertebrate lineage, implying a conserved function of these genes in neural crest development that arose at the origin of vertebrates(Meulemans and Bronner-Fraser,2002). Disruption of Tcfap2a gene function in mice revealed an essential role in neural crest survival, in addition to neural tube closure and middle ear development(Schorle et al., 1996; Zhang et al., 1996). We show that a zebrafish mutant in tfap2a displays similar defects in neural crest-derived tissues as those seen in the mouse Tcfap2a mutant. However, we have not detected similar defects in other ectodermal derivatives such as the neural tube, possibly reflecting differences in its development between fish and mouse.

In addition, we document a surprisingly early disruption of neural crest in low mutants, prior to migration, unlike that described in the mouse. Apoptosis in the neural crest of the pharyngeal arches was proposed as the cause of the mutant mouse phenotype, but no defects in neural crest migration or expression of early markers was described. We show that loss of tfap2a function in zebrafish disrupts early patterning of several populations of neural crest cells that form pigment, neurons and cartilage. Differences between the mouse and zebrafish mutants may be due to compensation by other AP2 genes in the mouse, such as Tcfap2b, which are expressed in the neural crest and may compensate for loss of Tcfap2a function. However, double mutant Tcfap2a/Tcfap2b mice still do not show early neural crest defects (T. Williams, personal communication). Alternatively, heterochronic differences in the timing of neural crest development of mouse compared with zebrafish might account for the observation that mouse Tcfap2a has a later function in pharyngeal arch neural crest development. We believe this unlikely, however, as such similar sets of neural crest derivatives are affected in mouse and zebrafish, all with a common origin in the hindbrain, implying an early requirement for tfap2a in their development. More detailed analyses of genes expressed in the neural crest in Tcfap2a mutant mice, should resolve these issues.

An important consequence of disruption of tfap2a function, is the loss of Hox group 2 gene expression in the hyoid and branchial arches. This requirement for tfap2a function in regulating Hox group 2 expression is conserved between mouse and zebrafish, and both mutants show disruption of the cartilages that in mammals give rise to the bones of the middle ear. Mice that are mutant for Tcfap2a display a range of phenotypes in which the middle ear bones are reduced or malformed (K. B. Avraham and K. P. Steel,personal communication). Human TFAP2A maps to chromosome 6p24.2 and a number of deafness syndromes have been mapped near to this location(Law et al., 1998). Furthermore, an individual has been described in whom TFAP2A is deleted, who has a reduced jaw, microphthalmia, proximally positioned thumbs and abnormal ear morphology (Davies et al., 1999). This suggests that Tfap2a has a highly conserved role in vertebrate development with many implications for human craniofacial and deafness syndromes.

We thank P. Ingham for support in the early stages of this work and P. Boardman for technical assistance. We also thank T. Williams and members of our laboratory for stimulating discussions and critical reading of the manuscript. R.G. and G.J.-R. thank Silke Rudolph-Geiger for technical assistance. We are grateful to J. Holzschuh, W. Driever, T. Williams and R. Cornell for communication of unpublished results. C. Amemiya provided arrayed PAC genomic libraries. We would like to thank the following people for probes:S. Johnson, J. Lister, V. Prince and D. Raible. This work was supported by the NIH (NS-41353, DE-13828), March of Dimes (1-FY01-198) and Pew Scholars Foundation (2615SC) to T.S., and by the German Human Genome Project (DHGP Grant 01 KW 9919).